The median number of high or critical vulnerabilities in cloud workloads increased fourfold between May and November this year, due to the impact of frontier AI, Datadog has claimed.
The cloud observability and security specialist revealed the findings in its State of Cloud Security 2026 published on November 10. It’s based on analysis of the cloud security posture of thousands of customer organizations around the world.
Datadog claimed that the median number of high/critical severity flaws in cloud workloads surged from four in May, when Anthropic’s Mythos model was released, to 14 in June and 16 in November.
The data was sourced from analysis of vulnerabilities in third-party libraries used by applications monitored through Datadog Code Security’s Software Composition Analysis (SCA) feature, the report noted.
“This data reveals that the increased number of vulnerabilities in the CVE database isn't a theoretical risk,” it added. “It affects an increasing number of real-world applications running in cloud environments, reinforcing the need for cloud security foundations that limit the blast radius of compromised applications.”
The report also highlighted the challenge of long-lived credentials; specifically related to cloud-hosted AI inference services and CI/CD pipelines.
Datadog warned that such credentials are consistently ranked as one of the most common causes of cloud breaches, with automated tools continuously scanning the web for weaknesses.
“Red-team frameworks like Pacu and Datadog’s own Pathrunner along with secret scanners like TruffleHog reliably find over-permissioned credentials and exposed keys at scale,” it said. “That same tooling is available to autonomous attackers, who enumerate for long-lived credentials first because they’re the highest-value, longest-lived prize in the environment.”
The report revealed that nearly half of organizations using Gemini use long-lived Gemini API keys, while the percentage of cloud users with access keys older than three years has reached nearly 30% for AWS IAM and Google Cloud Service Account users.
“Rather than static Bedrock or Gemini API keys, you should access AI inference services through the calling workload’s short-lived identity when supported (IAM role, managed identity, or service account) so the model is reached with the same federated credentials as other cloud resources and not a separate long-lived key,” Datadog explained.
Cloud Dependencies Expose Users to Machine-Speed Attacks
Supply-chain attacks are increasingly run autonomously and at machine speed, with hijacked dependencies one of the most popular routes to compromise, Datadog claimed.
Part of the challenge is the speed with which new dependency versions are adopted by users. If adopted automatically without proper vetting, they can result in the rapid compromise of cloud credentials and services.
Some 60% of organizations Datadog analyzed run versions that were released less than 24 hours earlier, with these versions appearing in more than a fifth (21%) of cloud applications.
The risk is amplified the more dependencies there are, and Datadog found that app workloads running in the cloud have on average 149 software dependencies. The figure across the average organization is 568.
“Organizations should implement dependency cooldowns to lower the risk of malicious dependencies reaching their cloud environments, because the community usually identifies and contains hijacked libraries within a few hours,” the report recommended.